Leadership style and group decision effectiveness: Correlational and behavioral tests of Fiedler's contingency model

Leadership style and group decision effectiveness: Correlational and behavioral tests of Fiedler's contingency model

0RGANIZATI0:NAL BEI:IAVIOR A:ND HCMAiN" I~ERFORMAI~ICE8, 347-362 (1972) ~.~Leadership Style and Group Decision Effectiveness: ~' Correlational and Be...

995KB Sizes 0 Downloads 22 Views

0RGANIZATI0:NAL BEI:IAVIOR A:ND HCMAiN" I~ERFORMAI~ICE8, 347-362 (1972)

~.~Leadership Style and Group Decision Effectiveness: ~' Correlational and Behavioral Tests of Fiedler's Contingency Model 1 MARSHALL SAStIKIN2

Wayne SLate University Experimental tests of Fiedler's "contingency model" of leadership effectiveness generally leave unanswered questions of precise criterion validity, due to limits on statistical significance of correlations based on small Ns. Seventy-five role-play groups were used in tests of one of the nine possible situations to which the model predicts. In this case, leaders with low LPC (Least-Preferred Co-Worker) scores ("task-oriented") should be superior to high LPC ("relationship-oriented") leaders. Two experiments (N = 53; N = 22) ~yielded r = - . 2 9 and -.19, respectively, between leaders' LPC and a group decision e'ffectiveness score. Overall, ra~ = - . 2 6 ( P ~ .05). The obtained confidence level closely approximates that which could be reasonably expected for the N used if the predicted correlation (-.33) were, in fact, the true population value. This, and the consistency of results between the two independent experiments, justifies the conclusion that a small but significant relationship exists between leadership style and group effectiveness for the situation investigated. The practical utility of this relationship was discussed. Some behavioral differences between high and low LPC leaders were also examined and discussed. Parts of the work reported here were conducted during the period of an N i l [ Traineeship Award (No. T-01-Ml[l1470-01). At that time facilities were provided to the authc~ under USPttS Grant No. Mtt-02704-09 to Dr. N. 1%. F. Maier. The latter portion of this research was performed while the author was graduate research assistant to Dr. Maier, as part of research supported by USPl[S Grant No. MI-I-02704-1L The research reported here is adapted from a doctoral dissertation by the author, conducted ai the University of Michigan. While the analyses and conclusions presented are the sole responsibility of the author, deepest appreciation is extended to Dr. Maier fQr his thoughtful assistance throughout this research. The author also thanks Dr. Anthony J. Reilly and Dr. Stanley E. Seashore for their support and aid. Dr. Michael J. I4avanagh was kind enough to offer constructive criticism and assistance in revision of an earlier draft of this report. • ~Requests for reprints should be sent to Marshall Sashkin, School of Business Administration, Department of Management, Wayne State University, Detroit, Michigan 48202. 347 Copyright © 1972 by Academic Press, Inc. All rights of reproduction in any form reserved.

348

MARSHALL SASI-IKIl~

Fiedler's (1967) "theory of leadership effectiveness" presents one of the clearest conceptions of the nature o f group leadership that has yet~ been offered. His position is fairly uncommon, in that it is conceptually based upon the idea that leadership style is a relatively stable personality attribute based within the motivational need patterns of the individual. While Fiedler cannot be categorized a s a traditional "trait" theorist, his position is certainly related to and possibly a derivative of such models. Thus, with trait theories in relative disfavor (e.g., Stogdill, 1948), it is not surprising that Fiedler's position has received vehement attacks (e.g., Graen, Alvares, Orris, & Martella, 1970; Graen, Orris, & Alvares, 1971), as well as having been subject to attempts to modify it in terms more congenial to current theorizing (e.g., Fishbein, Landy, & Hatch, 1969). Perhaps the single most significant issue concerning Fiedler's theory is its empirical validity. On the basis of much experimental research, in both field and laboratory settings, Fiedler (1967, 1968, 1971) concludes that a reasonable degree of validity has been established for his "contingency model." Others (e.g., Butterfield, 1968; Graen etal., 1970, 1971) challenge this. Since both Fiedler and his critics have, in their empirical tests of the contingency model, relied on correlations between the LPC (Least=Preferred Co-worker) measure of leadership style and some objective measure of group output, productivity, or effectiveness, considerable controversy has centered on the ways in which this statistic has been obtained and manipulated. It is not the intent of this paper to consider all of the problems of correlational measures, with regard to tests of the contingency model. We will 'be concerned specifically with the problem presented by the relation between sample size and level of statistical significance. The problem here is an outgrowth of the research style adopted by Fiedler, his colleagues, and his critics, and of the nature of the contingency model (which is the operational aspect of Fiedler's theory). Briefly, the contingency model provides for nine possible combinations of situational variables along a theoretically unidimensional scale of "favorablehess" of the situation for the exercise of leadership. The three dimensions of. group atmosphere, task structure, and leader's position power are operationally dichotomized such that eight combinations of these factors result, with a ninth added to describe situations of extreme unfavorableness (Fiedler, 1968). The problem of significance referred to previously arises from the desire of laboratory researchers to test the model in all its Variety and the problem faced by field experimenters of fig~ling many identical real-life groups. Thus, it has been common practice of both prop'onents and critics of the contingency model to devise o r ~ e n t i f y several groups in each of the eight or nine situation-types (or "cells").

LEADERSI-IIP S T Y L E AND GI~0TJP D]~ciSi0N EFFECTIVENESS

349

The theory predicts leadership effectiveness to be contingent upon the match between an individual's leadership style (as determined by the LPC instrument) and the particular situation-type the individual finds himself inl or is placed in. Thus, half of the experimental groups might be assigned leaders scoring high on the LPC (i.e., "relationship-oriented" leaders) while half receive leaders with low LPC scores ("task-oriented" leaders), or leaders might be randomly chosen and LPC scores obtained later. Obviously, the result is usually a small number of groups on which to base the correlation, since the number of available groups would first be apportioned into the eight or nine cells, and then, perhaps, further split by assigning high and low LPC leaders to half of the number in each cell. The number of groups in each cell is, as Gulch (1965) notes, related to the attainable level of significance of the correlational measure. According to MeHugh (1957) this relationship can be mathematically defined, if one specifies the desired confidence level and the tolerable deviation of the obtained correlation from that present in the population. The formula offered by McHugh (1957) is N = 3 + z2/d ~, where z is the normal deviate value of the desired confidence level and d is the deviation permissible. Accepting the 5% confidence limits as the minimal level of certainty desirable in an experimental study, and allowing for a permissible deviation of .10 for the correlation, the required sample size is 387 (Guion, 1965). With a larger tolerance for deviation, the required sample size decreases. For example, if we set .20 as the permissible deviation, N is reduced to 99. The research strategy of those concerned with the contingency model has generally resulted in samples of 8-12 groups per cell, sometimes even fewer. Significant and reliable correlations are unlikely on the basis of such low Ns. Fiedler (1967, 1971) attacks this problem indirectly, by obtaining correlations from a number of independent studies, each study involving a small number of groups per cell, and then calculating a median correlation for each cell of the model. Assuming a certain degree of consistency among the obtained correlations, such a strategy is by no means invalid. This strategy results in a broad pattern of empirical results which can be compared with theoretically based predictions. As Guion (1965) notes, "Much . . . practical significance . . . can be attached to results that turn out the same way in several repeated studies . . . [p. 126]." However, the direct question of precise criterion validity remains unanswered. The present study attempts to attack this last issue directly, by foregoing the desire to test the entire model and, instead, concentrating on only one of the situations defined by the model. With a reasonably large

3~0

:YIARSttALL SASttKIN

number of experimental groups, in one situation, the empirical validity of the model, for that situation, could be more adequately assessed in terms of correlational significance. A second issue with respect to the contingency model concerns the actual behaviors which might differentiate leaders with high LPC scores from low-scoring leaders. Two field studies, by Butterfield (1968) and by Reilly (1968), provided unexpected results in that low LPC leaders were behaviorally more relationship-oriented and less task-oriented than high LPC leaders. In each ease, however, these opposite-to-prediction behaviors occurred along with leadership effectiveness in the predicted direction. One might then suggest that the high or low LPC leader is effective when able to concentrate his leadership efforts on that aspect of the job situation which is not his primary concern. Gruenfeld, Rance, and Weissenberg (1969) found that in a laboratory experiment high LPC leaders were more accepting and exhibited more tension release behavior (both relationship-oriented behaviors) as compared with low LPC leaders, while the low LPC leaders were more dominant. High LPC leaders were also more effective, in general, on the experimental task. However, the experiment actually consisted of three, sequential, situations (as defined by the contingency model), two favoring low and one favoring high LPC leaders. Thus, these results are not necessarily inconsistent with Butterfield (1968) or Reilly (1968), since leadership effectiveness in the direction opposite to that predicted is then associated with behavior in the predicted direction, and the complementary half of the above explanation could be invoked. Fiedler (1967) has presented an argument and rationale similar to that cited; based on two of his early studies and a later validation study. Briefly, the leader is said to engage in behavior which is in accord with his primary pattern (task- or relationship-orientation) in unfavorable situations, but when his needs are being met or present no problem, as in favorable situations, he will engage in behaviors which conform to his secondary pattern. Thus, in a favorable situation a low LPC (taskoriented) leader would be expected to engage in relationship-oriented behaviors, while in an unfavorable situation the same leader would behave in a task-directed manner. The performance of the group is then dependent on which type of behavioral orientation is called for by the particular task situation. The studies cited above would seem to give some support to Fiedler's argument. However, despite complex data analyses (Fiedler, 1967; Reilly, 1968) it is still rather unclear what specific behaviors are associated with leadership style (LPC) and leadership effectiveness in a given situation. Maier and Sashkin (1971) have presented strong evidence with respect

LEADERSHIP STYLE AND GROUP DECISION EFFECTIVENESS

351

to two specific behaviors: the approach the leader takes in his initial contact with the group, and the extent to which he shares with the group information relevant to the task. In the present study, these two behaviors will be examined with respect to leaders' LPC scores and group effectiveness. In testing the contingency model other researchers have explored differences in how behaviors are used, rather than the absolute/requency of behaviors (Graham, 1968; Sample & Wilson, 1965). These studies have shown significant differences in the frequency distributions of task and relationship behaviors of high versus low LPC leaders, when there are no differences in the mean frequencies of the behaviors studied. In the present study, an attempt will be made to relate specific behaviors associated with effective leadership to leaders' LPC scores. Application of Graham's (1968) argument, concerning the variances of such behaviors for high versus low LPC leaders, will also be explored. The dependent variables for the present investigation are, then, measures of group performance, with regard to the first issue raised above, and measures of leader behavior, with respect to the second issue. METHOD

Cell, or situation-type, 3 of the contingency model was selected for the present investigation. According to Fiedler (1967), cell 3 is very favorable for the exercise of leadership. Cell 3 is defined by good leadermember relationships, unstructured task, and high leader position power. The contingency model predicts leaders with low LPC scores (taskoriented) to be more effective in this situation, as compared with high LPC (relationship-oriented) leaders. The decision to use a cell 3 situation was made for two reasons: (1) cell 3 has yielded the widest range of correlational results in past studies (Fiedler, 1967, p. 146), thus presenting a particular problem for the median-correlation approach; (2) a role play case which appeared to possess situational characteristics of cell 3 was available for use in the present study simply by administering the LPC instrument. A role play case was used because of the opportunity afforded by this method for tight situational control. Thus the particular case was chosen on grounds of utility as a test situation and, as noted, convenience. Case Used. The role play used was Maier's (1952) "Changing Work Procedures" (CWP) case. Briefly, the case involves a foreman and three workers, engaged in a discussion about changes in the work method used by this particular work group. The job is composed of three positions which the workers rotate hourly. Thus, each man works each lob position several times each day. A recent time study has, however, shown that pro-

352

MARSHALL SASHKIN

ductlon would increase significantly if each man were to work only at that job position he is best at. Although such an increase would mean an economic advantage for the workers they resist such a change for various reasons. Two of them find the rotation procedure makes the job less boring and the third is fearful of exposing his deficiencies to the foreman. Further, they all enjoy the group interaction which the rotation procedure permits. Examination of past results using this case shows it has proven sensi-. tire to several significant leadership factors, among them the attitude of the leader toward the group members (Maier & Hoffman, 1965), the approach of the leader (i.e., "selling" a solution versus posing a problem [Maier & tIoffman, 1964]) and his specific behaviors (Maier & Sashkin, 1971), and the degree of conflict present between the leader and group members (Hoffman, Harburg, & Maier, 1962). These factors seem quite relevant for the contingency model as they reflect leader-member relations (group atmosphere), task structure, and the leader's use of power in the group, which are the three operationally defined measures of "favorableness." The favorableness variables would seem well-defined in the CWP case. Examination of the workers' roles shows them to be positive or neutral toward the leader--certainly not negative toward him--thus defining the group atmosphere as moderately good, at least prior to the group interaction. The problem itself is relatively unstructured, as can be seen by examination of Shaw's criteria of task structure as discussed by Fiedler (1967, p. 28). The leader's position power is quite clear: he is direct supervisor of this work group and has final authority on questions of work procedure. These three factors clearly define the CWP as a cell 3 situation-type. In this case the contingency model predicts that leaders who derive need satisfactions from task accomplishment will prove more effective than leaders whose primary satisfactions are obtained through interpersonal relationships. Subjects. All subjects were university undergraduates, students in introductory psychology courses who participated in psychology experiments as part of their course requirement. Sixty four-person groups, composed of females only, participated in the first experiment in the fall of 1968. In the fall of 1969, the experiment was replicated with an additional 26 groups, composed of males and females. Data collection. At the conclusion of the role play discussion all subleers completed a brief questionnaire which included the following items: (1) a brief description of the solution adopted by the group; (2) the predicted efficiency of this solution (whether production would, under the solution agreed upon, go up, remain the same, or go down) ; (3) the ex-

LEADERSHIP STYLE A N D G R O U P DECISION EFFECTIVENESS

353

tent to which the leader shared his information concerning the problem with his subordinates (did not share any information; shared or told about some information; shared all the information); (4) leader's approach to the group (posed a problem; offered a Choice of alternatives; imposed his own solution). In addition subjects in the first experiment reporte~their degree of satisfaction with the leader on a five-point scale ranging from very satisfied (1) to very dissatisfied (5), and those i n t h e second study reported on the leader's helpfulness ("helped the group in reaching a decision"; "did not affect the decision reached"; "hindered the group in reaching a decision"). The L P C instrument, an 18-item semantic differentiaP measuring the favorableness with which the respondent views some individual in his past experience who he designates as his least-preferred co-worker, was completed by all subjects. Group products. Maier (1952) categorizes group solutions in the CWP case as "Old" (no change in work procedure), "l~ew" (adoption of the time study recommendation), and "Integrative." Solutions in the last category are considered to be of higher objective quality than the other two types, as Integrative solutions are creative products O]group discussion, avoiding either of the obvious alternatives (to change or not), and taking into account both the needs of workers and the desires of management for increased productivity. An example would be for each man to rotate back to his best job position more frequently, thus increasing the proportion of work time that each worker spends at his most productive iob, while retaining the advantages of the rotation system. Since Integrative solutions are creative, they vary widely as to specific details, but all possess these two basic factors of concern for the needs of workers and concern for management's desires. While quality of the solution is a major factor for the overall effectiveness of the decision, a second and in this case equally important factor is the acceptance the individual workers show for the particular group solution adopted (Niaier, 1963). Even a low quality solution could be highly effective if workers were committed enough to ma]ce it work well. Thus, an obiective measure of the effectiveness of the solution and, indirectly, of the effectiveness of the leader, must combine the two factors of quality and acceptance. This has been accomplished in the present study by scoring the type of solution as 1 = Old, 2 -- New, or 3 = Integrative, and then multiplying this quality score by a group acceptance score. The latter was derived from the previously mentioned workers' re: 8Fiedler has traditionally used an 8-point scale: The 6-pointl scale used here was adapted from l~eilly (1968). Comparison of the LPC distributions obtained in the present investigation with that obtained by Fiedler (1967) shows no significant differences. In both cases LPC scores appear to conform to a normul~distributi0n.

354

MARSHALL SASI-IKIN

ports of future productivity. The responses o f increased production, no change, or decreased production were scored + 1, 0, and - 1 , respectively. The raw group acceptance score was the sum of the three individual scores and thus ranged from + 3 to - 3 . Addition of a constant ( = 4) to each group score gave a positive range of 7-1. Since decision effectiveness (DE) is defined as t h e product of the quality and the acceptance measures, the potential range of D E is from 1 to 21. 4 RESULTS

Definition o] situation. Of the three variables that define a contingency model situation, only one seems at all likely to vary from its original definition in the CWP ease, namely, the group atmosphere. For the 60 groups in the first experiment group scores of satisfaction with the leader were obtained simply by summing =the three individual reports and dividing by 3. The overall mean for all 60 groups was 1.96, slightly better than "moderately satisfied," which corresponds quite well to the degree of favorableness of group atmosphere that was intended. It is also possible that the group atmosphere was different for groups with high versus those with low LPC leaders. However, comparison of mean satisfaction-with-leader scores for the 28 high versus 32 low LPC leader grouPS showed no difference at all (1.97 vs 1.96). The LPC of the leader had no particular impact on the satisfaction of the group members with the leader. A final question is whether all groups were "moderately satisfied." Examination of the data showed seven groups at or below the scalar midpoint (3 = "neither satisfied nor dissatisfied"). These groups were eliminated from the following data analysis. The four groups below the mean can be considered as examples of cell 7 situations (moderately poor leader-member relations, unstructured task, high leader position power). These data will not, however, be examined, since no statistical meaning can be attached to a correlation involving an N of four. For the 26 groups in the replication study no measure of satisfaction with leader was obtained. However, group members did report on the leaders'helpfulness in reaching a decision ("helped the group in reaching a decision"; "did not affect the decision reached"; "hindered the group in reaching a decision"). This measure may also be considered as an indicator of leader-member relations, since groups reporting the leader as primarily a hindrance would be expected to be less satisfied with their 4Several possible weightings were tried o u t for the components of D E . The use of the'differen~ Weights for the quality and acceptance measures h a d no appreciable effect upon the correlation between leaders' L P C scores and :DE. For further details, see Sashkin (1970). ~ .

LEADERSHIP STYLE AND GROUP DECISION EFFECTIVENESS

355

leader than groups reporting him as primarily helpful. E x a m i n a t i o n - o f data from 13 groups in a recent modification of this same experimental situation showed t h a t the leaders' perception of group atmosphere (measured on ten semantic differential scales and considered b y Fiedler [1967] to be the most relevant measure of l e a d e r - m e m b e r relations) was significantly related to the group members' reports of leaders' helpfulness (r = .52, P < .05). Thus, it seems empirically justified to consider the group's helpfulness rating of the leader as a measure of leader-member relationships. Scoring the cited "helpfulness" alternatives as 3 (helpful), 2 (no effect), and 1 (hindered), and summing to obtain a group score for each group, showed four groups in which the leader was neither a help nor a hindrance (group score = 6). I n the remaining 22 groups, the leader was invariably reported to be of help (groups scores of 7-9). T h e four groups mentioned were not included in the following analysis. H a v i n g eliminated from the data analysis the seven groups w i t h low or indifferent satisfaction-with-leader scores in the first experiment, and the four groups not reporting the leader as helpful in the second experiment, it seems reasonable to consider the remaining 53 groups in the first and 22 groups in the second experiment as celt 3 situations. Leadership style and group :decision effectiveness. Since leaders were chosen for their high or low L P C scores in the first experiment (resultihg in a bimodal distribution on this variable), a biserial correlation was most appropriate to examine the degree of association between leaders' L P C scores and group decision effectiveness (DE) (Guilford, 1965, p. 321). T a b l e 1 shows t h a t for the 53 groups in the first experiment r~ = - . 2 9 (P < .05). For the second experiment, r = - . 1 9 . While this result does not approach statistical significance, it confirms and thus strengthens the confidence one m a y place in the prior outcome. Since the second experiment does seem to present an adequate replication of the first, in result as

TABLE 1 I~ELATIONSHIPS BETWEEN LEADERS' LPC SCORES AND GROUP DECISION EFFECTIVENESS SCORES r

~]xperiment

N

(LPC-DE)

I~ II Combined

53 22 75

.29b .19 - . 26b

~ The correlation is a biserial r. b p < .05. .

~

356

1VIARSItALL SASHKIN

well as method, it is justifiable to pool the correlations and obtain an average r, such that the sample size is as large as possible, in line with the primary objective of a large-scale contingency model test. Our earlier discussion permits a rough estimation of the required sample size, with regard to attaining statistically significant results. If a deviation of about .20 from the true population value of the correlation is acceptable, the actual sample size available, N = 75, is not too far from that calculated as necessary from McHugh's (1957) formula (N = 99). The obtained value of r = - . 2 6 reaches an acceptable level of significance (P < .05). Indeed, given the present sample size, this level of significance is even greater than one might expect, in light of McHugh's analysis of required sample size and taking into consideration the desired level of significance and limit on deviation which were noted above and used to estimate the required N. Throughout the present analysis, the concern has been whether the obtained correlation differs significantly from zero. In these terms, it has already been noted that the present result (r--- -.26) does reach an acceptable level of statistical significance. However, since the contingency model makes a specific prediction based on the mean of correlations obtained by Fiedler (1967), it is equally relevant to ask whether the obtained correlation is significantly different from that predicted. The mean correlation obtained by Fiedler (1967) for studies concerned with cell 3 is -.33, which is treated as a specific prediction from his model. The result obtained here does not differ significantly from this contingency model prediction (z = .119, n.s.).5 Thus, the preseng result can be considered consistent with Fiedler's (1967) earlier data and with his prediction based on those data, as well as being significantly different from zero.

Leadership style and leader behaviors. Table 2 presents correlations between leaders' LPC scores and behaviors of leaders (as reported by group members). Of particular interest are the correlations between leaders' LPC scores and information sharing behavior, and between D E and information sharing. In the former case r = .19, while for the latter two variables r = .10, which seems rather low considering the strong relationship between solution quality and extent of information sharing established by h/~aier and Sashkin (1971) in a study using this experimental situation. However, it is clear that the small advantage already This statistical test is not, strictly speaking, legitimate, as Fiedler's prediction is a mean of rank-order, not product-moment, correlations. The result is presented here since there seems to be no other method for testing the obtained difference and Fiedler (personal communication) believes his result would be unchanged by using r, rather than rho (which he used for convenience of calculation with small Ns).

LEADERSHIP

STYLE

AND

GROUP

TABLE

DECISION

2

RELATIONSHIPS BETWEEN GROUP DECISION EFFECTIVENESS SCORES LEADERS' LPC SCORES, AND REPORTED LEADER BEHAVIORSa

Experiment

N

I

53

II

22

Combined

Note.--Correlations

75

Behavioral measure Information sharing Leaders' approach Information sharing Leaders' approach Information sharing Leaders ~ approach

between LPC

357

EFFECTIVENESS

AND

r

r

(LPCbehavior)

(DEbehavior)

,24 b .13 .06 .18 .19 .14

.08 .24 b .14 ,27 .i0 .25 b

and Behavioral measures in Expt. I are biserial

T~S.

bp ~ .05. noted for low LPC leaders (in terms of DE) cannot be due to these leaders' information sharing behavior, since that behavior is positively related to leaders' LPC. The data on the approach of the leader to the group (extent to which the leader's approach was problem-oriented, rather than solutionoriented) is equally unhelpful in understanding the slight advantage of low LPC leaders. Maier and Sashkin (19.71) found that a more problemoriented approach was strongly associated with high quality group decisions. Here, the correlation between leaders' LPC scores and leaders' approach scores is .14, while for the leaders' approach and DE r is .25. The latter correlation makes even more puzzling the positive relationship between leaders' LPC scores and leaders' behavior (degree of problemorientation). One possible explanation can be seen in results obtained by Graham (1968) and Sample and Wilson (1965). Sample and Wilson found that high and low LPC scoring leaders did not differ in frequencies of task and socioemotional behaviors (Bales, 1950) but did differ as to when they employed these behaviors. Graham (1968) suggests that circumstances which evoke these behaviors are different: leaders with high LPC scores engage more generally in socioemotional (relationship-oriented) behaviors, while those with low LPC scores use these behaviors to motivate or reward task performance. Thus, socioemotional behaviors should occur with less variation among high LPC leaders as compared to low LPC leaders, although the overall frequency of such behaviors might be the same for both groups. Similarly, low LPC leaders should show more consistency in task-related behaviors than high LPC leaders.

358

""

MARSHALL SASI-It(IN

TABLE 3 MEANS AND V£RIANCES OF HIGH VERSUS L o w L P C LEADERS FOR THE BEHAVIORS OF INFORMATION SHARING AND LEADERS' APPROACH TO THE GROUP

High L P C leaders

Low L P C leaders

Behavior

N

M

~2

N

5/[

~

F

Information sharing Leaders' approach

35 35

4.71 6.03

3.51 1.97

40 40

4.05 5.78

1.39 2.88

2.52 a 1.46

~ P < .02.

Information sharing seems to be, in ~the CWP case, primarily a taskrelated behavior. The mos~ frequent reason given for not sharing the information is concern for the subordinates' feelings, since the data clearly show the superiority of two of the workers over the third, and identify each man's strengths and weaknesses. The leader's approach to the group is basically a matter of establishing a relationship with the group members, and can be considered as a socioemotional type of behavior. Following Graham's (1968) argument, then, one might expect greater variance in information sharing by high (as compared with l o w ) L P C leaders, while the reverse would be anticipated with regard to leaders' approach. By dividing the leaders into two groups, along the mean of leaders' LPC scores, the variances for the two behaviors can be examined for high as compared with low LPC leaders. Table 3 presents these results. As can be seen, leaders with high LPC scores show significantly greater variance for information sharing behavior, although the means are not significantly different. While low LPC leaders do show greater variance in approach to the group, this difference is not statistically significant. The means, again, do not differ significantly. DISCUSSION AND CONCLUSIONS

': Given the agreement between Fiedler's and the present results, the confirmation of results via the replication of the first by the second of the two experiments reported here (despite differences in sample populations and in the :distributions of leaders' LPC scores), and the significance of the large sample result, it seems appropriate to conclude that there does exist a consisten~ and significant, though not very large, relationship between leadership style (as measured by the leaders' LPC scores) and group decision effectiveness (as defined above) for type 3 situations (as defined by the contingency model). There remains the question of what specific behaviors differentiate

LEADERSHIP STYLE AND GROUP DECISION EFFECTIVENESS

359

high from low~LPC leaders, particularly with respect to facilitation of group performance. Measures of two behaviors, earlier shown to be strongly related to the quality of group solutions for the particular experimental situation used here, were obviously not responsible for the slight superiority of leaders with lower LPC scores, since both behavioral measures showed positive (although not significant) relationships with leaders' LPC scores as well as positive relationships with D E (Table 2). Earlier studies have yielded behavioral differences for high versus low LPC leaders which are seemingly in accord with expectations derived from Fiedler's model (Gruenfeld, Rance, & Weissenberg, 1969), as well as differences which appear to be quite opposite to such expectations (Butterfield, 1968; Reilly, 1968). One possible reconciliation and interpretation of these results was given above, and Fiedler (1967) was noted to have presented a similar argument. With respect to the present results , since Fiedler (1971) now asserts that it is only under conditions of extremely poor situational favorableness that leader behaviors differ strongly in terms of contingency model expectations, perhaps the present weak relationships are not surprising. In any case, two specific leader behaviors which do seem strongly related to the quality of group decision making have been shown here not to be those used by low LPC leaders to obtain better group products. What such behaviors might be, if there are any identifiable, is unanswered. Obviously, this is an area for further research. On the positive side, the data here do support Graham's (1968) conclusion that while high and low LPC leaders do not differ in terms of mean frequencies of task or socioemotional behaviors, there is a difference in the way high and low LPC leaders use these behaviors, as shown by the different degrees of variance~ ia these behaviors for high versus low LPC leaders (Table 3). The data here actually complement Graham's findings. His most significant difference between behavior variances of high and low LPC leaders concerned relationship-oriented behavior, while tile present results show a significant difference for a task-directed behavior. Granting a degree of predictive validity, by LPC scores, of leadership effectiveness, which seems to be on the order of - . 2 0 to - . 3 0 for cell 3; one is left with the question of the practical utility of such a finding. Fiedler (1967, 1968) has suggested an approach he refers to as "organizational engineering," by which leaders would be placed in situations most suited to their particular leadership styles as measured by the LPC instrument. The value of such an administrative procedure certainly seems questionable for cell 3 situations, considering the degree of relationship demonstrated here, with leadership style accounting for at most

360

:gJIARSttALL SASIIKIN

10% of the variance in group performance. There also remains the fact of the wide range of correlational relationships reported for this situation type by Fiedler (1967). Organizational engineering is likely to be even less useful in certain other circumstances, e.g., situation-type 7, where the median correlation reported by Fiedler is .05. Fiedler's (1968) data do, however, show considerably larger median correlations for several of the situations (specifically, cells l, 2, 8, and 9--or 5-A---of the contingency model), although large-scale tests to determine more precisely the degree of validity present in these cases have yet to be performed. Still another problem was dealt with by Duffy, Kavanagh, 1VIacKinney, Wolins, and Lyons (1970), who found in a field study that the model was practically useless when the criterion measure of leadership effectiveness was a rating by the leader's supervisor, rather than the "hard" criteria usually used by Fiedler (1967). Since ratings are frequently used in industrial situations, Duffy et al., concluded that the contingency model would not be particularly useful in such cases. The generally weak correlations between LPC and effectiveness which have been obtained in laboratory studies are considered by t0iedler (1971) to be due to limitations of the studies, since field data generally show stronger relationships. The design and control of the present laboratory study did yield a result very close to the contingency model prediction. However, if Fiedler's assertion is generally true, then accurate validation studies on the contingency model will require improved laboratory designs and methods (beginning with the determination of what specifically is inadequate in these experiments) or exhaustive, largescale field experiments. Even the large sample field study approach may be inadequate. Schein (1971) recently noted that field researchers are not now and are unlikely to be in the future able to build adequate controls into their studies, and has proposed a search for new research models. In the meantime, rather than attempting to redesign or "engineer" an organization entirely on the basis of contingency model predictions, it would seem more fruitful to apply the model in specific, carefully chosen instances, where chances for successful results are known to be greatest. It may ultimately be found that the contingency model can be an effective tool when used selectively, even though it is not a very effective panacea. REFERENCES BALES, 1%. F. Interaction process analysis. Cambridge, Massachusetts: AddisonWesley, 1950.

LEADERSHIP STYLE AND GROUP DECISION EFFECTIVENESS

361

BUTTERFIELD,D. A. An integrative approach to the study of leadership effectiveness in organizations. Unpublished doctoral dissertation, The University of Michigan, 1968. DVFFY, J. F., KAVA~AaH,M. J., MAcKIzc~Ev, A. C., WOLINS, L., & LYozcs, T: F. A field extension of Fiedler's contingency model. Paper presented at the meeting of the Midwestern Psychological Association, Cincinnati, 1900. FIEDLER, F. E. A theory o] leadership effectiveness. New York: McGraw-Hill, 1967. FIEI)LER, F. E. Personality and situational determinants of leadership effectiveness. In D. Cartwright and A. Zander (Eds.), Group dynamics. (3rd ed.) New York: Harper & Row, 1968. FIEDLER, F. :E. Validation and extension of the contingency model of leadership effectiveness: A review of empirical findings. Psychological Bulletin, 1971, 76, 128-148. FISHBEIN, M., LANDY, E., • HATCH, G. Some determinants of an individual's esteem for his least preferred co-workers: An attitudinal analysis. Human Relations, 1959, 22, 173-188. GRAEN, G., ALYARES,K. M., ORRIS, J. lB., ~¢ MARTELLA,J. A. Contingency model of leadership effectiveness: Antecedent and evidential results. Psychological Bulletin, 1970, 74, 285-296. GRAE~, G., ORRIS, J. B., & ALVARES,K. M. Contingency model of leadership effectiveness: Some experimental results. Journal of Applied Psychology, 1971, 55, 196-201. GRAHAM, W. I(. Description of leader behavior and evaluation of leaders as a function of LPC. Personnel Psychology, 1968, 21, 457-464. GRUENFELD, L., RANC'E, D. E., & WEISSEIqBERG,P. The behavior of task-oriented (low LPC) and socially oriented (high LPC) leaders under several conditions of social support. Journal of Social Psychology, 1969, 79, 99-107. GUILFORD,J. P. Fundamental statistics in psychology and education. (4th ed.) New York: McGraw-Hill, 1965. Gvloz~, 1%. M. Personnel testing. New York: McGraw-Hill, 1965. HO~FMAN, L. R., t~aB~raG, E., & MAItre, N. R. F. Differences and disagreement as factors in creative group problem solving. Journal o/Applied Psychology, 1962, 64, 206-214. MCHVGH, 1%. B. Determining sample size in validation research. Educational and Psychological Measurement, 1957, 17, 136-141. MAma, N. R. F. Principles o] human relations. New'York: Wiley, 1952. MAma, N. R. F. Problem-solving discussions and conferences. New York: McGrawHill, 1963. MAIER, N. 1%. F., & HOFFMAN, L. 1%. Acceptance and quality of solutions as related to leaders' attitudes toward disagreement in group problem solving. Journal of Applied Behavioral Science, 1965, 1, 373-386. MAIER, N. 1%. F., & SAS~KIN, M. Specific leadership behaviors that promote problem solving. Personnel Psychology, 1971, 24, 35--44. REILI,Y, A. J., III. The effects of different leadership styles on group performance: A field experiment. Ames, Iowa: Industrial Relations Center, Iowa State University, 1968. SAMPLE, J. A., & WILSON, T. 1%. Leader behavior, group productivity, and rating of least preferred co-workers. Journal of Personality and Social Psychology, 1965, 1, 266-270. SASItKIN, M. Supervisory leadership in problem solving groups: Experimental tests

362

MARSHALL SASHKIN

of Fred Fiedler's "Theory of Leadership Effectiveness" in ~he laboratory using role play methods. Unpublished doctoral dissertation, The University of Michigan, 1970. SCI~EII% E. H. Organizational psychology: Problems and prospects. Paper presented at the meeting of the American Psychological Association, Washington, DC, 1971. STOaDII,L, R. M. Personal factors associated with leadership: A survey of the literature. Journal o] Psychology, 1948, 25, 35-71. RECEIVED: N o v e m b e r 20, 1971